1. Introduction¶
This report from the ACCESS Metrics team contains information about the utilization and performance of the resources from the Resource Provider PSC for the dates 2025-04-01 through 2025-06-30. The data in this report come from ACCESS XDMoD via the ACCESS Central Database as well as data provided directly to the ACCESS Metrics team by the Resource Provider. You can contact the ACCESS Metrics team regarding the data in this report by submitting a ticket. Below is a summary of the data for each active resource during this time period.
Compute Resources¶
Bridges 2 RM¶
There were 544 active projects (projects that ran at least one job).
There were 1,600 active users (users who ran at least one job).
The top 5 parent fields of science in terms of number of active projects were:
Engineering and Technology (115 projects)
Biological Sciences (99 projects)
Chemical Sciences (87 projects)
Physical Sciences (81 projects)
Computer and Information Sciences (76 projects)
Bridges 2 GPU¶
There were 319 active projects (projects that ran at least one job).
There were 1,010 active users (users who ran at least one job).
The top 5 parent fields of science in terms of number of active projects were:
Computer and Information Sciences (134 projects)
Biological Sciences (52 projects)
Engineering and Technology (41 projects)
Chemical Sciences (22 projects)
Physical Sciences (21 projects)
Bridges 2 EM¶
There were 37 active projects (projects that ran at least one job).
There were 39 active users (users who ran at least one job).
The top 5 parent fields of science in terms of number of active projects were:
Biological Sciences (9 projects)
Physical Sciences (9 projects)
Chemical Sciences (5 projects)
Engineering and Technology (5 projects)
Computer and Information Sciences (4 projects)
PSC Neocortex¶
There were 4 active projects (projects that ran at least one job).
There were 11 active users (users who ran at least one job).
The top 5 parent fields of science in terms of number of active projects were:
Computer and Information Sciences (3 projects)
Other (1 projects)
Biological Sciences (0 projects)
Bridges2 GPU AI¶
There were 2 active projects (projects that ran at least one job).
There were 2 active users (users who ran at least one job).
The top 5 parent fields of science in terms of number of active projects were:
Computer and Information Sciences (1 projects)
Earth and Environmental Sciences (1 projects)
Physical Sciences (0 projects)
Engineering and Technology (0 projects)
Mathematics (0 projects)
PSC Neocortex SDFlex¶
There were 3 active projects (projects that ran at least one job).
There were 24 active users (users who ran at least one job).
The top 5 parent fields of science in terms of number of active projects were:
Computer and Information Sciences (2 projects)
Other (1 projects)
Biological Sciences (0 projects)
2. Resource Utilization¶
This section shows utilization information for ACCESS-allocated compute resources. To faciliate comparison across different types of compute resources (such as CPU or GPU), usage is presented in ACCESS Credit Equivalents. Each compute resource has a conversion factor between a resource service unit and an ACCESS Credit. For example, one ACCESS Credit Equivalent is equal to 1 CPU core hour on SDSC Expanse (which uses AMD EPYC 7742 processors).
The data here are based on information reported to ACCESS by the RP via the Usage Reporting API.
Figure 2.1 Usage of Bridges 2 RM from 2023-04-01 to 2025-06-30. Usage is shown converted to ACCESS Credit Equivalents. The red solid line shows the hypothetical upper bound for credit usage based on the number of compute elements on the resource. The green dotted line shows the mean value of the ACCESS Credits that the resource is obliged to make available for use by ACCESS allocated jobs.
Figure 2.2 Usage of Bridges 2 GPU from 2023-04-01 to 2025-06-30. Usage is shown converted to ACCESS Credit Equivalents. The red solid line shows the hypothetical upper bound for credit usage based on the number of compute elements on the resource. The green dotted line shows the mean value of the ACCESS Credits that the resource is obliged to make available for use by ACCESS allocated jobs.
Figure 2.3 Usage of Bridges 2 EM from 2023-04-01 to 2025-06-30. Usage is shown converted to ACCESS Credit Equivalents. The red solid line shows the hypothetical upper bound for credit usage based on the number of compute elements on the resource. The green dotted line shows the mean value of the ACCESS Credits that the resource is obliged to make available for use by ACCESS allocated jobs.
Figure 2.4 Usage of PSC Neocortex from 2023-04-01 to 2025-06-30. Usage is shown converted to ACCESS Credit Equivalents. The red solid line shows the hypothetical upper bound for credit usage based on the number of compute elements on the resource. The green dotted line shows the mean value of the ACCESS Credits that the resource is obliged to make available for use by ACCESS allocated jobs.
Figure 2.5 Usage of Bridges2 GPU AI from 2023-04-01 to 2025-06-30. Usage is shown converted to ACCESS Credit Equivalents. The red solid line shows the hypothetical upper bound for credit usage based on the number of compute elements on the resource. The green dotted line shows the mean value of the ACCESS Credits that the resource is obliged to make available for use by ACCESS allocated jobs.
Figure 2.6 Usage of PSC Neocortex SDFlex from 2023-04-01 to 2025-06-30. Usage is shown converted to ACCESS Credit Equivalents. The red solid line shows the hypothetical upper bound for credit usage based on the number of compute elements on the resource. The green dotted line shows the mean value of the ACCESS Credits that the resource is obliged to make available for use by ACCESS allocated jobs.
3. Active Projects¶
This section shows information about ACCESS projects running on the resources. A project is counted as active in a time period if there was resource usage charged to the project. Projects that had an allocation on the resource but no usage are not included in these data.
Figure 3.1 Number of Active Projects by Resource by Quarter, Last Two Years.
By Parent Field of Science¶
This section shows the breakdown of active projects by the parent field of science of the project. The field of science information is provided by the project's PI when they apply for or update their project information. The parent field of science shown in this data is obtained from each project's primary field of science.
Figure 3.2 Number of Active Projects on Bridges 2 RM by Parent Field of Science by Quarter, Last Two Years.
Figure 3.3 Number of Active Projects on Bridges 2 GPU by Parent Field of Science by Quarter, Last Two Years.
Figure 3.4 Number of Active Projects on Bridges 2 EM by Parent Field of Science by Quarter, Last Two Years.
Figure 3.5 Number of Active Projects on PSC Neocortex by Parent Field of Science by Quarter, Last Two Years.
Figure 3.6 Number of Active Projects on Bridges2 GPU AI by Parent Field of Science by Quarter, Last Two Years.
Figure 3.7 Number of Active Projects on PSC Neocortex SDFlex by Parent Field of Science by Quarter, Last Two Years.
4. Active Users¶
This section shows information about active users. The active user metrics are based on the batch job or virtual machine owners. These metrics do not include counts of people who were indirectly associated with the compute usage such as science gateway end-users or users interacting with web services run inside VMs.
Figure 4.1 Number of Active Users by Resource by Quarter, Last Two Years.
By Parent Field of Science¶
This section shows the counts of users by the parent field of science of the ACCESS project that was charged.
Figure 4.2 Number of Active Users on Bridges 2 RM by Parent Field of Science by Quarter, Last Two Years.
Figure 4.3 Number of Active Users on Bridges 2 GPU by Parent Field of Science by Quarter, Last Two Years.
Figure 4.4 Number of Active Users on Bridges 2 EM by Parent Field of Science by Quarter, Last Two Years.
Figure 4.5 Number of Active Users on PSC Neocortex by Parent Field of Science by Quarter, Last Two Years.
Figure 4.6 Number of Active Users on Bridges2 GPU AI by Parent Field of Science by Quarter, Last Two Years.
Figure 4.7 Number of Active Users on PSC Neocortex SDFlex by Parent Field of Science by Quarter, Last Two Years.
By Academic Status¶
This section shows the counts of users by the Academic Status value from their ACCESS profile. This value is self reported by users when they first register for an ACCESS account. A value of Unknown means that they did not complete their profile information.
Figure 4.8 Number of Active Users on Bridges 2 RM by User Academic Status by Quarter, Last Two Years.
Figure 4.9 Number of Active Users on Bridges 2 GPU by User Academic Status by Quarter, Last Two Years.
Figure 4.10 Number of Active Users on Bridges 2 EM by User Academic Status by Quarter, Last Two Years.
Figure 4.11 Number of Active Users on PSC Neocortex by User Academic Status by Quarter, Last Two Years.
Figure 4.12 Number of Active Users on Bridges2 GPU AI by User Academic Status by Quarter, Last Two Years.
Figure 4.13 Number of Active Users on PSC Neocortex SDFlex by User Academic Status by Quarter, Last Two Years.
5. Open OnDemand Usage¶
This section shows information about the usage of Open OnDemand obtained from web server logs sent directly to the ACCESS Metrics team from the RP. A user is counted as using Open OnDemand if they have logged into Open OnDemand on a resource and they have an ACCESS affiliated account on the resource.
Figure 5.1 Number of Open OnDemand Users by Resource by Quarter, Last Two Years.
6. Science Gateway Usage¶
This section shows information about compute usage via Science Gateways that have a compute allocation and a community account. These data come from the information reported by ACCESS integrated Science Gateways via the reporting API. Not all gateways use the API, so the user counts represent a lower bound of the total number of people who are making use of compute via a gateway.
Figure 6.1 Number of Science Gateway Users by Resource by Quarter, Last Two Years.
7. Network Flows¶
Network flows and Globus Transfer information for ACCESS RPs are tracked via NetSage for ACCESS.
8. Job Performance¶
Job Efficiency data in ACCESS XDMoD are obtained from performance data collected on the compute nodes combined with accounting information from job scheduler logs and the ACCESS Usage API.
Additional data are available in the Efficiency tab in ACCESS XDMoD (some data may only be visible if you have the Center Staff or Center Director role; submit a ticket to the ACCESS Metrics team to request this role if you do not have it).
CPU Usage for CPU-based Compute Resources¶
This section shows CPU usage data for CPU-based compute resources. The CPU User value is the ratio of CPU time elapsed where the operating system is in user mode to total CPU time for the cores that the job was assigned.
Figure 8.1 Scatter plot showing the average CPU Usage weighted by CPU hour against CPU Hours for jobs by User on Bridges 2 RM. All jobs with performance data between 2025-04-01 and 2025-06-30 are included. Data points in the top right of the chart indicate users who have high resource usage with low average utilization.
Figure 8.2 Bar chart showing CPU Usage binned by CPU User Value on Bridges 2 RM for compute jobs between 2025-04-01 and 2025-06-30. A value of NA indicates that no performance data was available for the job. This typically happens with short running jobs or when there is an issue with the data collection on the compute node.
Figure 8.3 Scatter plot showing the average CPU Usage weighted by CPU hour against CPU Hours for jobs by User on Bridges 2 EM. All jobs with performance data between 2025-04-01 and 2025-06-30 are included. Data points in the top right of the chart indicate users who have high resource usage with low average utilization.
Figure 8.4 Bar chart showing CPU Usage binned by CPU User Value on Bridges 2 EM for compute jobs between 2025-04-01 and 2025-06-30. A value of NA indicates that no performance data was available for the job. This typically happens with short running jobs or when there is an issue with the data collection on the compute node.
GPU Usage for GPU-based Compute Resources¶
This section shows GPU usage data for GPU-based compute resources. The GPU usage is the value reported by the GPU driver software running on the compute nodes and indicates the proportion of time where the GPU was executing instructions.
Figure 8.5 Bar chart showing GPU Usage binned by GPU Usage Value on Bridges 2 GPU for compute jobs between 2025-04-01 and 2025-06-30. A value of NA indicates that no performance data was available for the job. This typically happens with short running jobs or when there is an issue with the data collection on the compute node.
Figure 8.6 Bar chart showing GPU Usage binned by GPU Usage Value on Bridges2 GPU AI for compute jobs between 2025-04-01 and 2025-06-30. A value of NA indicates that no performance data was available for the job. This typically happens with short running jobs or when there is an issue with the data collection on the compute node.
Document Version History¶
| Version | Date | Notes |
|---|---|---|
| 1 | 2025-07-17 | Initial Version. |
XDMoD Data Analytics Framework 1.1.0